Neural network model image reconstruction method and device applied to financial service system
By generating convolutional neural networks built with adversarial networks and multi-scale pyramid structures, the problem of insufficient image reconstruction accuracy in financial service systems is solved, and higher-precision image reconstruction is achieved.
Patent Information
- Application Number
- CN202510306640.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-14
- Publication Date
- 2025-07-18
Smart Images
Figure CN120338025A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of neural network model image reconstruction for financial service systems. Specifically, it relates to a method, device, computer-readable storage medium, and electronic device for neural network model image reconstruction applied to financial service systems. Background Art
[0002] With the development of technology and to assist the efficient development of business, the financial industry has introduced more and more high-precision devices and intelligent systems, such as face recognition, monitoring, and intelligent scanning of materials in outlets. The image information that the financial system needs to process has shown an explosive growth. Depending on the performance of the shooting device, the quality of the shooting environment, and the source of the image, the initial images we obtain may be unclear. At this time, we need to use image super-resolution reconstruction technology to adjust the images to meet some scenarios with high requirements for image detail accuracy. The reconstruction algorithms of traditional methods, such as interpolation, filtering, denoising, and least squares method, have certain limitations, with low reconstruction accuracy or slow reconstruction speed. Summary of the Invention
[0003] The main purpose of the present application is to provide a method, device, computer-readable storage medium, and electronic device for neural network model image reconstruction applied to financial service systems, so as to at least solve the problem of low reconstruction accuracy in traditional image reconstruction methods for financial service systems.
[0004] To achieve the above object, according to one aspect of the present application, there is provided a method for neural network model image reconstruction applied to financial service systems, including: obtaining a training data set of the financial service system, where the training data set includes model input images and target output images corresponding to the model input images; constructing a conditional generative adversarial network model, and training the conditional generative adversarial network model with the training data set to obtain an image reconstruction model, where the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed using a CNN structure; obtaining an image to be reconstructed, and inputting the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
[0005] Optionally, before constructing the generator of the conditional generative adversarial network model using a generative adversarial network and a multi-scale pyramid structure, the method further includes: constructing the multi-scale pyramid structure using atrous convolutions with multiple different dilation rates to perform multiple samplings on the same feature map.
[0006] Optionally, the generator comprises an input module, a first preset number of cascaded residual modules, a second preset number of Residual-ASPP modules, and an output module. Among them, the input module is composed of a third preset number of convolutional layers, and each convolutional layer of the input module is connected with a BN layer and a Leaky ReLU activation function. The first preset number of cascaded residual modules are connected behind the input module. The Residual-ASPP module is used to combine the inter-layer skip connections in the residual module. The output module is composed of a convolutional layer and an upsampling convolutional layer.
[0007] Optionally, the Residual-ASPP module includes a fourth preset number of dilated convolutional channels and a merging layer. The merging layer is used to perform feature fusion processing on the features output by the fourth preset number of dilated convolutional channels.
[0008] Optionally, during the process of training the conditional generative adversarial network model using the training data set, the method further includes: using the discriminator of the conditional generative adversarial network model to determine whether the model output image output by the generator is accurate according to the target output image in the training data set; when the model output image output by the generator is accurate, determining that the conditional generative adversarial network model has completed training.
[0009] Optionally, during the process of training the conditional generative adversarial network model using the training data set, the method further includes: calculating the loss function of the conditional generative adversarial network model, where the loss function includes mean squared error and cross entropy; according to the loss function, using the backpropagation algorithm to perform iterative optimization processing on the model parameters of the conditional generative adversarial network model.
[0010] Optionally, after obtaining the training data set of the financial service system, the method further includes: performing data preprocessing operations on the training data set, where the data preprocessing operations include: image normalization operation, image standardization operation, image enhancement operation, image data denoising operation, and image data augmentation operation.
[0011] According to another aspect of the present application, there is provided an image reconstruction device for a neural network model applied to a financial service system, including: an acquisition unit configured to acquire a training data set of the financial service system, wherein the training data set includes a model input image and a target output image corresponding to the model input image; a construction unit configured to construct a conditional generative adversarial network model, and train the conditional generative adversarial network model by using the training data set to obtain an image reconstruction model, wherein the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed by using a CNN structure; an input unit configured to acquire an image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
[0012] According to still another aspect of the present application, there is provided a computer-readable storage medium, where the computer-readable storage medium includes a stored program, and when the program runs, it controls the device where the computer-readable storage medium is located to execute any one of the above-mentioned image reconstruction methods for a neural network model applied to a financial service system.
[0013] According to yet another aspect of the present application, there is provided an electronic device, including: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include those for executing any one of the above-mentioned image reconstruction methods for a neural network model applied to a financial service system.
[0014] By applying the technical solution of the present application, a training data set of a financial service system is acquired, wherein the training data set includes a model input image and a target output image corresponding to the model input image; a conditional generative adversarial network model is constructed, and the conditional generative adversarial network model is trained by using the training data set to obtain an image reconstruction model, wherein the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed by using a CNN structure; an image to be reconstructed is acquired, and the image to be reconstructed is input into the image reconstruction model to obtain a reconstructed image. By combining a generative adversarial structure and a multi-scale pyramid structure to build a convolutional neural network for image reconstruction in a financial service system, the problem of low reconstruction accuracy existing in traditional image reconstruction methods for financial service systems is solved. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] The accompanying drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and the descriptions thereof are used to explain this application and do not constitute an improper limitation to this application. In the drawings:
[0016] Figure 1 A hardware structure block diagram of a mobile terminal for implementing a neural network model image reconstruction method applied to a financial service system according to an embodiment of this application is shown;
[0017] Figure 2 A schematic flowchart of a neural network model image reconstruction method applied to a financial service system according to an embodiment of this application is shown;
[0018] Figure 3 A schematic diagram of a generative adversarial network structure according to an embodiment of this application is shown;
[0019] Figure 4 A schematic diagram of the process of dilated convolution according to an embodiment of this application is shown;
[0020] Figure 5 A schematic diagram of an atrous spatial pyramid pooling (ASPP) structure according to an embodiment of this application is shown;
[0021] Figure 6 A schematic diagram of the generator structure of an image reconstruction model according to an embodiment of this application is shown;
[0022] Figure 7 A structure block diagram of a neural network model image reconstruction device applied to a financial service system according to an embodiment of this application is shown.
[0023] Among them, the above-mentioned drawings include the following reference numerals:
[0024] 102, a processor; 104, a memory; 106, a transmission device; 108, an input / output device. Detailed implementation manners
[0025] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments may be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.
[0026] In order to enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.
[0027] It should be noted that the terms "first", "second", etc. in the specification, claims and above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances, so as to implement the embodiments of the present application described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.
[0028] As introduced in the background art, the traditional image reconstruction method of the existing financial service system has a low reconstruction accuracy. To solve the problem of low reconstruction accuracy of the traditional image reconstruction method of the financial service system, the embodiments of the present application provide a neural network model image reconstruction method, device, computer-readable storage medium and electronic device applied to the financial service system.
[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.
[0030] The method embodiments provided in the embodiments of the present application can be executed on a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 is a hardware structure block diagram of a mobile terminal of a neural network model image reconstruction method applied to a financial service system according to an embodiment of the present invention. As Figure 1 shown, the mobile terminal may include one or more ( Figure 1 only one is shown in Figure 1 a processor 102 (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Among them, the above-mentioned mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown in Figure 1 is only schematic and does not limit the structure of the above-mentioned mobile terminal. For example, the mobile terminal may further include more or fewer components than
[0031] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the neural network model image reconstruction method applied to the financial service system in the embodiments of the present invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, the above-mentioned method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the mobile terminal through a network. Examples of the above-mentioned network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include the wireless network provided by the communication provider of the mobile terminal. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0032] In this embodiment, a neural network model image reconstruction method applied to a financial service system running on a mobile terminal, a computer terminal, or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.
[0033] Figure 2 It is a flowchart of the neural network model image reconstruction method applied to the financial service system according to the embodiments of the present application. As Figure 2 shown, the method includes the following steps:
[0034] Step S201, obtaining a training data set of the financial service system, where the above-mentioned training data set includes model input images and target output images corresponding to the above-mentioned model input images;
[0035] Among them, the financial service system includes banks, securities companies, insurance companies, trust companies, fund companies, fintech companies, etc. These institutions provide various financial services, including savings, loans, investments, insurance, payments, settlements, etc.
[0036] Step S202: Construct a conditional generative adversarial network model, and train the conditional generative adversarial network model with the above training dataset to obtain an image reconstruction model. Among them, the conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed by using a CNN structure;
[0037] Among them, the conditional generative adversarial network (CGAN) is a machine learning model composed of a generator network and a discriminator network. The two compete with each other through adversarial training. The generator network is responsible for generating data samples, while the discriminator network is responsible for determining whether the generated data samples are real data or generated data. By continuously optimizing the parameters of the two networks, the generator network can finally generate realistic data samples, thus realizing the generation and synthesis of data.
[0038] Step S203: Obtain the image to be reconstructed, and input the image to be reconstructed into the above image reconstruction model to obtain a reconstructed image.
[0039] In this embodiment, a training dataset of the financial service system is obtained, where the training dataset includes model input images and target output images corresponding to the model input images; a conditional generative adversarial network model is constructed, and the conditional generative adversarial network model is trained with the training dataset to obtain an image reconstruction model. Among them, the conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed by using a CNN structure; the image to be reconstructed is obtained, and the image to be reconstructed is input into the image reconstruction model to obtain a reconstructed image. By combining the generative adversarial structure and the multi-scale pyramid structure to build a convolutional neural network for image reconstruction in the financial service system, the problem of low reconstruction accuracy existing in the traditional image reconstruction method of the financial service system is solved.
[0040] In the specific implementation process, before constructing the generator of the conditional generative adversarial network model by using a generative adversarial network and a multi-scale pyramid structure, the method further includes: building the multi-scale pyramid structure by using atrous convolutions with multiple different dilation rates to perform multiple samplings on the same feature map.
[0041] This method uses atrous convolutions with different dilation rates to build a pyramid structure, performs multiple samplings on the same feature map, and then performs serial fusion. This structure helps to capture features of different scales simultaneously and effectively avoids obtaining redundant information.
[0042] Specifically, the above generator consists of an input module, a first preset number of cascaded residual modules, a second preset number of Residual-ASPP modules, and an output module. Among them, the above input module consists of a third preset number of convolutional layers, and after each of the above convolutional layers of the above input module, a BN layer and a Leaky ReLU activation function are connected. The first preset number of cascaded above residual modules are connected behind the above input module. The above Residual-ASPP module is used to combine the skip connections between layers in the above residual module. The above output module consists of a convolutional layer and a transposed convolutional layer.
[0043] Among them, the first preset number can be set to 6, the second preset number can be set to 2, and the third preset number can be set to 2.
[0044] The generator of this method contains an input module, which consists of two convolutional layers, and after each convolutional layer, there is a BN layer and a Leaky ReLU activation function; then there are six cascaded residual modules, the structure of which is shown in the dotted box on the right, and two Residual-ASPP modules are embedded therein. Finally, there is an output module, which consists of two convolutional layers plus a transposed convolutional layer.
[0045] More specifically, the above Residual-ASPP module includes a fourth preset number of dilated convolutional channels and a merging layer, and the above merging layer is used to perform feature fusion processing on the features output by the fourth preset number of the above dilated convolutional channels.
[0046] Among them, the fourth preset number can be set to 3.
[0047] Each Residual-ASPP module of this method contains three dilated convolutional channels with dilation rates of 6, 12, and 24 respectively, followed by a merging layer to fuse the features of the three dilated convolutional channels.
[0048] Furthermore, in the process of training the above conditional generative adversarial network model using the above training dataset, the above method further includes: using the discriminator of the above conditional generative adversarial network model to determine whether the model output image output by the above generator is accurate according to the above target output image in the above training dataset; in the case where the model output image output by the above generator is accurate, determining that the above conditional generative adversarial network model has completed training.
[0049] In the process of training the model of this method, when the model output image output by the model is compared with the real target output image and they are consistent, it is considered that the conditional generative adversarial network model has completed training.
[0050] Furthermore, in the process of training the conditional generative adversarial network model using the above training dataset, the above method further includes: calculating the loss function of the conditional generative adversarial network model, where the loss function includes mean squared error and cross entropy; and iteratively optimizing the model parameters of the conditional generative adversarial network model according to the loss function by using the backpropagation algorithm.
[0051] This method uses the loss function and the backpropagation algorithm to constrain the process and effect of model learning, continuously reducing the difference between the model output image and the target output image.
[0052] Specifically, after obtaining the training dataset of the financial service system, the above method further includes: performing data preprocessing operations on the training dataset, where the data preprocessing operations include: image normalization operation, image standardization operation, image enhancement operation, image data denoising operation, and image data augmentation operation.
[0053] The specific data preprocessing operations of this method include:
[0054] 1. Image normalization: Scale the image pixel values to a unified range, usually [0,1] or [-1,1].
[0055] 2. Image size adjustment: Adjust the image to a unified size, usually by cropping or scaling to a fixed size.
[0056] 3. Image enhancement operation: Perform enhancement operations on the image, such as rotation, flipping, brightness adjustment, etc., to increase data diversity.
[0057] 4. Image data augmentation operation: Generate new training samples by randomly transforming the original image, thereby increasing the data volume.
[0058] 5. Image standardization operation: Standardize the image so that its mean is 0 and variance is 1.
[0059] 6. Data augmentation: Perform some transformations on the image, such as rotation, translation, scaling, etc., to increase data diversity.
[0060] 7. Data balancing: Balance the data to keep the number of samples in each category relatively balanced.
[0061] 8. Image data denoising operation: Remove the noise in the image to improve the quality and accuracy of the image.
[0062] In order to enable those skilled in the art to more clearly understand the technical solution of this application, the implementation process of the neural network model image reconstruction method applied to the financial service system of this application will be described in detail below with specific embodiments.
[0063] In the financial industry, the usage scenarios of images are becoming increasingly diverse. Whether it is the face recognition system in branches or the financial service system for processing customer information, key information needs to be extracted from images, such as customer portrait recognition, information recognition of submitted materials, etc. When the quality of the original images we obtain is not high, the accuracy of information extraction from the images will be reduced, such as low face recognition rate, incorrect reading of material information, etc. Compared with text information, the difficulty in processing image information lies in its large and complex data volume. Images are composed of pixel points, and each pixel point contains color and brightness information. Moreover, the details and structures in images are complex and variable. In addition, images are visual representations of the real world. Considering their quality and fidelity, it is necessary to maintain the accuracy of image details and colors, which also increases the processing difficulty. What we are researching is how to quickly transform blurred images into clear images using reconstruction algorithms with as little original information as possible and maximize the restoration of reality. This field is also known as image super-resolution reconstruction.
[0064] The most commonly used traditional methods are interpolation method and regularization method. The interpolation method assumes that adjacent pixel points have similar pixel values and uses existing pixel values to estimate missing pixel values. This method has high computational efficiency but still results in blurred images and is suitable for scenarios with limited computing resources. The regularization method introduces additional constraints or prior information during the reconstruction process and uses the statistical characteristics of images to improve the reconstruction accuracy. However, when the structure of the image content is too complex, it will lead to over-smoothing. Therefore, the present invention uses deep learning to enhance the ability to restore image details while maximizing the reconstruction resolution and not reducing the reconstruction speed.
[0065] This embodiment relates to a specific neural network model image reconstruction method applied to a financial service system. A convolutional neural network for image reconstruction in the financial service system is built by combining a generative adversarial structure and a multi-scale pyramid structure. The specific contents are as follows:
[0066] First, in the generative adversarial structure, we use a Conditional Generative Adversarial Networks (CGAN), and the structure is as Figure 3 shown, which consists of a generator (G) and a discriminator (D).
[0067] Generative adversarial networks can directly generate the required images from random noise through training. Their advantage is that they do not require prior information and can generate sharper and clearer samples than other models. However, sometimes this method is too free and uncontrollable. One solution is to add some constraints to the generative adversarial network and add prior information such as labels or image features to the input, thus conditional generative adversarial networks emerged. In the present invention, the required prior information is the lossy image after compression.
[0068] Secondly, for the multi-scale pyramid structure, we use the Atrous Spatial Pyramid Pooling (ASPP) structure. This structure utilizes the advantages of atrous convolution. By adding holes to the convolutional kernel, it not only retains the computational cost of small convolutional kernels but also can expand the receptive field and retain the internal data structure, such as Figure 4 shown.
[0069] The pyramid structure is built using atrous convolutions with different dilation rates to sample the same feature map multiple times and then concatenate and fuse them. This structure helps to capture features of different scales simultaneously and effectively avoids obtaining redundant information. This is ASPP, as Figure 5 shown.
[0070] Based on the above two structures, we constructed a generator as shown in Figure 6 and creatively proposed the module. This module combines ASPP with the inter-layer skip connection in the residual network to solve the problem of training difficulties brought by network depth and enables the network to capture image information of different scales without hindering the feed-forward of information.
[0071] As Figure 6 shown, the generator contains an input module composed of two convolutional layers, each followed by a BN layer and a Leaky ReLU activation function; then there are six cascaded residual modules, the structure of which is shown in the right dashed box, and two Residual-ASPP modules are embedded, the structure of which is shown in the left dashed box. For each Residual-ASPP module, it contains three atrous convolution channels with dilation rates of 6, 12, and 24 respectively, followed by a merging layer to fuse the features of the three channels; finally, there is an output module composed of two convolutional layers plus a restoration convolutional layer. The discriminator uses a relatively classic CNN structure.
[0072] In the embodiments of the present application, a convolutional neural network is built by combining the generative adversarial structure and the multi-scale pyramid structure for image reconstruction in the financial service system, solving the problem of low reconstruction accuracy existing in the traditional image reconstruction method of the financial service system.
[0073] The embodiments of the present application further provide a neural network model image reconstruction device applied to a financial service system. It should be noted that the neural network model image reconstruction device applied to the financial service system in the embodiments of the present application can be used to execute the neural network model image reconstruction method applied to the financial service system provided by the embodiments of the present application. The device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.
[0074] The following introduces the neural network model image reconstruction device applied to the financial service system provided by the embodiments of the present application.
[0075] Figure 7 is a schematic diagram of the neural network model image reconstruction device applied to the financial service system according to the embodiments of the present application. As Figure 7 shown, the device includes:
[0076] An acquisition unit 71, configured to acquire a training data set of the financial service system, where the training data set includes model input images and target output images corresponding to the model input images;
[0077] A construction unit 72, configured to construct a conditional generative adversarial network model, and train the conditional generative adversarial network model with the training data set to obtain an image reconstruction model, where the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed by using a CNN structure;
[0078] An input unit 73, configured to acquire an image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
[0079] In this embodiment, an acquisition unit is configured to acquire a training data set of a financial service system, where the training data set includes model input images and target output images corresponding to the model input images; a construction unit is configured to construct a conditional generative adversarial network model, and train the conditional generative adversarial network model by using the training data set to obtain an image reconstruction model, where the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed by using a CNN structure; an input unit is configured to acquire an image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image. By combining a generative adversarial structure and a multi-scale pyramid structure to build a convolutional neural network for image reconstruction in a financial service system, the problem of low reconstruction accuracy existing in the traditional image reconstruction method of the financial service system is solved.
[0080] As an optional solution, the device further includes a building unit configured to build the multi-scale pyramid structure by using atrous convolutions with multiple different dilation rates to sample the same feature map multiple times.
[0081] An optional solution is that the generator includes an input module, a first preset number of cascaded residual modules, a second preset number of Residual-ASPP modules, and an output module. The input module is composed of a third preset number of convolutional layers, and a BN layer and a Leaky ReLU activation function are connected after each of the convolutional layers in the input module. The first preset number of cascaded residual modules are connected after the input module. The Residual-ASPP module is used to combine the inter-layer skip connections in the residual module. The output module is composed of a convolutional layer and a transposed convolutional layer.
[0082] An optional solution is that the Residual-ASPP module includes a fourth preset number of atrous convolution channels and a merging layer, and the merging layer is used to perform feature fusion processing on the features output by the fourth preset number of atrous convolution channels.
[0083] As an optional solution, the device further includes a first determination unit and a second determination unit. The first determination unit is configured to, during the process of training the conditional generative adversarial network model by using the training data set, use the discriminator of the conditional generative adversarial network model to determine whether the model output image output by the generator is accurate according to the target output image in the training data set. The second determination unit is configured to determine that the conditional generative adversarial network model is completed when the model output image output by the generator is accurate.
[0084] An alternative solution is that the device further includes a calculation unit and an iterative optimization processing unit. The calculation unit is used to calculate the loss function of the conditional generative adversarial network model during the process of training the conditional generative adversarial network model with the above training data set, where the loss function includes mean square error and cross entropy. The iterative optimization processing unit is used to perform iterative optimization processing on the model parameters of the conditional generative adversarial network model according to the loss function by using the backpropagation algorithm.
[0085] An alternative solution is that the device further includes a preprocessing unit, which is used to perform data preprocessing operations on the above training data set after obtaining the training data set of the financial service system. The data preprocessing operations include: image normalization operation, image standardization operation, image enhancement operation, image data denoising operation, and image data augmentation operation.
[0086] The above neural network model image reconstruction device applied to the financial service system includes a processor and a memory. The above acquisition unit, construction unit, input unit, etc. are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions. The above modules are all located in the same processor; or, the above modules are respectively located in different processors in any combination form.
[0087] The processor contains a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be set, and by adjusting the kernel parameters, the problem of low reconstruction accuracy existing in the traditional image reconstruction method of the financial service system can be solved.
[0088] The memory may include non-permanent memory in computer-readable media, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM). The memory includes at least one storage chip.
[0089] An embodiment of the present invention provides a computer-readable storage medium. The computer-readable storage medium includes a stored program, where when the program runs, it controls the device where the computer-readable storage medium is located to execute the above neural network model image reconstruction method applied to the financial service system.
[0090] Specifically, the neural network model image reconstruction method applied to the financial service system includes:
[0091] Step S201, obtaining a training data set of the financial service system, where the training data set includes model input images and target output images corresponding to the model input images;
[0092] Step S202, construct a conditional generative adversarial network model, and train the conditional generative adversarial network model with the above training data set to obtain an image reconstruction model. The conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed by using a CNN structure;
[0093] Step S203, obtain the image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
[0094] An embodiment of the present invention provides a processor for running a program. When the program runs, it executes the above neural network model image reconstruction method applied to a financial service system.
[0095] Specifically, the neural network model image reconstruction method applied to a financial service system includes:
[0096] Step S201, obtain a training data set of a financial service system, where the training data set includes a model input image and a target output image corresponding to the model input image;
[0097] Step S202, construct a conditional generative adversarial network model, and train the conditional generative adversarial network model with the above training data set to obtain an image reconstruction model. The conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed by using a CNN structure;
[0098] Step S203, obtain the image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
[0099] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements at least the following steps:
[0100] Step S201, obtain a training data set of a financial service system, where the training data set includes a model input image and a target output image corresponding to the model input image;
[0101] Step S202, construct a conditional generative adversarial network model, and use the above training dataset to train the conditional generative adversarial network model to obtain an image reconstruction model. Among them, the conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed using a CNN structure;
[0102] Step S203, obtain the image to be reconstructed, and input the image to be reconstructed into the above image reconstruction model to obtain a reconstructed image.
[0103] The device in this article can be a server, a PC, a PAD, a mobile phone, etc.
[0104] This application also provides a computer program product, which is suitable for executing a program initialized with at least the following method steps when executed on a data processing device:
[0105] Step S201, obtain a training dataset of a financial service system, where the training dataset includes model input images and target output images corresponding to the model input images;
[0106] Step S202, construct a conditional generative adversarial network model, and use the above training dataset to train the conditional generative adversarial network model to obtain an image reconstruction model. Among them, the conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed using a CNN structure;
[0107] Step S203, obtain the image to be reconstructed, and input the image to be reconstructed into the above image reconstruction model to obtain a reconstructed image.
[0108] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. They can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order than here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.
[0109] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.
[0110] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.
[0111] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.
[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in Figure 1 one or more of the processes or multiple processes and / or blocks Figure 1 one or more of the blocks or multiple blocks.
[0113] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.
[0114] The memory may include non-permanent memory in the computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.
[0115] A computer-readable medium includes permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette tapes, magnetic disk storage, or other magnetic storage devices, or any other non-transitory medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.
[0116] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising the element.
[0117] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:
[0118] 1) A neural network model image reconstruction method applied to a financial service system of the present application includes: obtaining a training data set of the financial service system, where the training data set includes model input images and target output images corresponding to the model input images; constructing a conditional generative adversarial network model, and training the conditional generative adversarial network model with the training data set to obtain an image reconstruction model. The conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed using a CNN structure; obtaining an image to be reconstructed, and inputting the image to be reconstructed into the image reconstruction model to obtain a reconstructed image. It solves the problem of low reconstruction accuracy existing in the traditional image reconstruction method of the financial service system.
[0119] 2) An image reconstruction device of a neural network model applied to a financial service system according to the present application includes: an acquisition unit for acquiring a training data set of the financial service system, where the training data set includes a model input image and a target output image corresponding to the model input image; a construction unit for constructing a conditional generative adversarial network model and training the conditional generative adversarial network model with the training data set to obtain an image reconstruction model, where the conditional generative adversarial network model includes a generator and a discriminator, the generator is constructed by using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure is an atrous spatial pyramid pooling structure, and the discriminator is constructed by using a CNN structure; and an input unit for acquiring an image to be reconstructed and inputting the image to be reconstructed into the image reconstruction model to obtain a reconstructed image. It solves the problem that the traditional image reconstruction method of the financial service system has a low reconstruction accuracy.
[0120] The foregoing are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.
Claims
1. A neural network model image reconstruction method applied to a financial service system, characterized in that, Including: Obtain a training data set of a financial service system, where the training data set includes model input images and target output images corresponding to the model input images; Construct a conditional generative adversarial network model, and use the training data set to train the conditional generative adversarial network model to obtain an image reconstruction model. The conditional generative adversarial network model includes a generator and a discriminator. The generator is constructed using a generative adversarial network and a multi-scale pyramid structure, and the multi-scale pyramid structure is an atrous spatial pyramid pooling structure. The discriminator is constructed using a CNN structure; Obtain an image to be reconstructed, and input the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
2. The method according to claim 1, wherein Before constructing the generator of the conditional generative adversarial network model using a generative adversarial network and a multi-scale pyramid structure, the method further includes: Build the multi-scale pyramid structure using atrous convolutions with multiple different dilation rates to sample the same feature map multiple times.
3. The method according to claim 1, wherein The generator consists of an input module, a first preset number of cascaded residual modules, a second preset number of Residual-ASPP modules, and an output module. The input module consists of a third preset number of convolutional layers, and after each convolutional layer of the input module, a BN layer and a Leaky ReLU activation function are connected. The first preset number of cascaded residual modules are connected behind the input module. The Residual-ASPP module is used to combine the inter-layer skip connections in the residual module. The output module consists of a convolutional layer and a deconvolutional layer.
4. The method according to claim 3, wherein The Residual-ASPP module includes a fourth preset number of atrous convolution channels and a merging layer, and the merging layer is used to perform feature fusion processing on the features output by the fourth preset number of atrous convolution channels.
5. The method according to claim 1, wherein During the process of training the conditional generative adversarial network model using the training data set, the method further includes: Use the discriminator of the conditional generative adversarial network model to determine whether the model output image output by the generator is accurate according to the target output image in the training data set; When the model output image output by the generator is accurate, determine that the conditional generative adversarial network model has completed training.
6. The method according to claim 1, wherein During the process of training the conditional generative adversarial network model using the training data set, the method further includes: Calculate the loss function of the conditional generative adversarial network model, where the loss function includes mean squared error and cross entropy; According to the loss function, use the backpropagation algorithm to perform iterative optimization processing on the model parameters of the conditional generative adversarial network model.
7. The method according to claim 1, wherein After obtaining the training data set of the financial service system, the method further includes: Perform data preprocessing operations on the training data set, where the data preprocessing operations include: image normalization operation, image standardization operation, image enhancement operation, image data noise reduction operation, and image data augmentation operation.
8. An image reconstruction device of a neural network model applied to a financial service system, characterized in that, Comprising: An acquisition unit for acquiring a training data set of a financial service system, where the training data set includes model input images and target output images corresponding to the model input images; A construction unit for constructing a conditional generative adversarial network model, training the conditional generative adversarial network model using the training data set to obtain an image reconstruction model, where the conditional generative adversarial network model includes a generator and a discriminator, constructing the generator using a generative adversarial network and a multi-scale pyramid structure, the multi-scale pyramid structure being an atrous spatial pyramid pooling structure, and constructing the discriminator using a CNN structure; An input unit for acquiring an image to be reconstructed and inputting the image to be reconstructed into the image reconstruction model to obtain a reconstructed image.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, where when the program runs, it controls the device where the computer-readable storage medium is located to execute the neural network model image reconstruction method applied to a financial service system according to any one of claims 1 to 7.
10. An electronic device, characterized in that, Comprising: One or more processors, a memory, and one or more programs, where the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include those for executing the neural network model image reconstruction method applied to a financial service system according to any one of claims 1 to 7.